Persons with dementia display frequent problematic vocalizations and other dementia-related behaviors that perplex, dismay, or annoy others in the environment. These behaviors often lead to negative consequences for the patient, and caregivers lack theoretically-based interventions to address them. Based on an extensive literature review, our interactive research project group has reconceptualized them as need-based dementia-compromised behaviors (NDBs) and we now propose to examine variables that predict which persons with dementia are most at risk for displaying them and under what conditions they are most likely to occur. The purpose of our study, one of a trio of interactive projects is to develop a preliminary explanatory model of problematic vocalizations that will serve as a basis for empirical testing of targeted interventions that ultimately will improve the quality of life for persons with dementia and their caregivers. Our specific aims are to: 1) develop a model of background (relatively stable) and proximal (more changeable) variables affecting problematic vocalizations in persons with dementia; and 2) assess convergent validity of our measures of problematic vocalizations; with the verbally agitated behaviors of the Cohen- Mansfield Agitation Inventory. Using a multivariate cross-sectional descriptive design, we will enroll a cluster sample of 236 residents from nine randomly selected nursing homes. Measures of background variables include physical examination and instrumentation with participants and a consenting family informant. Measures of proximal variables include chart review, questioning of participants, assessment of environmental ambiance, and mechanical measurements of participants' movement, environmental light, noise, and heat index. Measures of problematic vocalizations will come from videotapes and audiotapes made during 14 ten-minute observations. Hierarchical linear modeling will be used to test a model of problematic vocalizations. The resulting preliminary model will be further tested in two independent samples obtained for the accompanying projects. Further, data from a core set of background and proximal factors common across the three interactive projects will be combined to develop a model of factors operative in producing NDBs in general.